Nature Biotechnology 24, 358 - 362 (2006)
Published online: 12 February 2006; | doi:10.1038/nbt1187
An in vitro fluorescence screen to identify antivirals that disrupt hepatitis B virus capsid assemblyStephen J Stray1, Jennifer M Johnson1, Benjamin G Kopek1, 2
& Adam Zlotnick11 Department of Biochemistry and Molecular Biology, University of Oklahoma Health Sciences Center, 975 NE 10th St. BRC 456, Oklahoma City, Oklahoma 73104, USA. 2 Present address: McArdle Laboratory for Cancer Research, University of Wisconsin-Madison, Madison, Wisconsin 53706, USA.
Correspondence should be addressed to Adam Zlotnick adam-zlotnick@ouhsc.edu Virus assembly has not been routinely targeted in the development of antiviral drugs, in part because of the lack of tractable methods for screening in vitro. We have developed an in vitro assay of hepatitis B virus (HBV) capsid assembly, based on fluorescence quenching of dye-labeled capsid protein, for testing potential inhibitors. This assay is adaptable to high-throughput screening and can identify small-molecule inhibitors of virus assembly that prevent, inappropriately accelerate and/or misdirect capsid formation to yield aberrant particles. An in vitro primary screen has the advantage of identifying promising lead compounds affecting assembly without the requirement that they be taken up by cells in culture and be nontoxic. Our approach may facilitate the identification of antivirals targeting viruses other than HBV, such as avian influenza and HIV.
HBV is a serious public health problem worldwide1, with more than 400 million people chronically infected by this small enveloped DNA virus. HBV has an icosahedral core, comprising viral nucleic acid and reverse transcriptase, enclosed in a protein capsid composed of 240 copies of the capsid (or core) protein2. Core assembly is essential for replication, as DNA synthesis occurs exclusively within the core particle3. We have studied HBV capsid assembly extensively in vitro using the N-terminal assembly domain (residues 1–149) of strain adyw capsid protein expressed in Escherichia coli4. This protein lacks the C-terminal nucleic acid binding domain (residues 150–183), dispensable for assembly of empty capsids5. The HBV capsid protein is a dimer in solution4,
6. In vitro assembly of HBV capsid protein depends on protein concentration, NaCl concentration, pH and temperature4,
7,
8, and is probably regulated allosterically8,
9,
10. Assembly is nucleated by a trimer of core-protein dimers, followed by rapid addition of subsequent dimers7. A network of weak intersubunit contacts holds HBV capsids together8, a property common to many virus systems11. Capsids persist even under unfavorable conditions because disassembly and reassembly reactions compete, leading to a kinetic barrier to dissociation (hysteresis)12.
Heteroaryldihydropyrimidines (HAPs)13,
14, initially discovered in a tissue culture–based screen, have recently been shown to inhibit HBV replication by perturbing capsid assembly10,
15. In vitro assembly is accelerated in the presence of HAP-1, a representative HAP compound, leading to aberrant particles (sheets and tubes) at higher HAP-1 concentrations. This suggests that HAPs act as synthetic allosteric activators. HAP-1 also reorganized core protein from preassembled capsids into noncapsid polymers, presumably by interaction of HAP-1 with dimers freed during capsid 'breathing', the transitory breaking of individual intersubunit bonds10.
To more readily identify assembly-directed molecules, we describe an in vitro system for assessing the ability of small molecules to alter HBV capsid assembly. We demonstrate the sensitivity of this assay using small molecules that prevent normal HBV capsid assembly, either by inhibiting capsid formation (urea12) or by accelerating and misdirecting assembly (HAP-1 (ref. 10)).
Capsid assembly causes core-protein C termini to be brought close together, suggesting that distance-sensitive probes such as the fluorescent BODIPY dyes could be attached at the C terminus and undergo self-quenching when assembly occurs. Cryo–electron microscopy studies have shown that the C termini of C150 are located on the interior of the assembled capsid and are in close proximity16. This has been supported in an X-ray crystal structure (1QGT)17, although density halted short of the C termini. Measured from the last modeled amino acids in 1QGT, distances between C termini of adjacent subunits range from 12.3 Å to 14.8 Å. Distances across the fivefold and sixfold vertices range from 17.3 Å to 21.6 Å; the remaining six to seven disordered residues could easily allow substituents at the C termini to collide across a vertex. By comparison, within a dimer, the last ordered C-terminal residues are about 50 Å apart.
For fluorescent labeling, we chose mutant C150 (ref. 16), where all wild-type cysteines have been mutated to alanines and a unique cysteine was added to the C terminus, allowing facile labeling with cysteine-reactive agents. All of the cysteine residues of the HBV core protein are dispensable for core assembly, DNA replication and particle production18. C150 was nearly quantitatively labeled with BODIPY-FL maleimide (C150BO). Dimeric C150BO was highly fluorescent, but fluorescence was markedly reduced when C150BO was assembled (Fig. 1).
 | | |
Self-quenching of BODIPY dyes occurs by formation of a ground-state dimer19. This nonfluorescent dimer can also act as an acceptor for fluorescence resonance energy transfer, further quenching fluorescence20. The absorbance spectrum of the unassembled C150BO dimer is typical for the BODIPY-FL monomer, but the assembled C150BO shows a reduction in the absorbance of the BODIPY-FL peak at 504 nm and a shoulder at 477 nm, qualitatively consistent with the presence of BODIPY-FL dimers19 (Supplementary Fig. 1 online).
We previously characterized assembly of HBV capsid protein using real-time 90° light scattering and size-exclusion chromatography (SEC)8,
9,
10,
21. Here we tested assembly of C150BO by light scattering using a range of NaCl concentrations (Fig. 1a). In parallel experiments, we tested C150BO for assembly by light scattering and BODIPY-FL fluorescence. The rate and extent of assembly increased with increasing NaCl concentration, as seen with the wild type, whereas fluorescence decreased (Fig. 1b).
Comparison of assembly kinetics by light scattering and fluorescence (Fig. 1a,b) shows that the curves were essentially identical when rescaled (Fig. 1c), even during the earliest phases of the reaction. Most data points lie on the diagonal when plotting light scattering versus fluorescence over 100 s to 600 s (Fig. 1d). The slopes of the fluorescence versus light scattering curves were the same for [NaCl]  0.75 M, indicating a strong correlation between the signals. To compare fluorescence quenching to the extent of assembly at equilibrium, we measured fluorescence and light scattering for overnight assembly reactions (conditions as for Fig. 1); subsequently, each sample was subjected to SEC (Fig. 2). C150BO assembly was more complete than wild-type assembly under the same conditions8; thus association energy is slightly higher for C150 and C150BO (Supplementary Table 1 and Supplementary Fig. 2 online). BODIPY-FL labeling did not significantly affect assembly. Fluorescence, light scattering and SEC measurements of assembly at equilibrium were self-consistent. Assembly increased over the range 0–0.75 M NaCl, whereas fluorescence decreased proportionally over this same range owing to quenching of the dye-labeled molecules.
| | |
SEC of C150BO assembly reactions showed intermediates not usually seen in wild-type assembly reactions; intermediates were less evident at [NaCl] 0.5 M (Supplementary Fig. 2 online). Similar SEC profiles were seen with unlabeled C150, suggesting that the mutations in C150 differentially affect nucleation of assembly, especially at high NaCl concentrations, leading to kinetic traps as predicted when association energy is increased (as we observe) and/or nucleation no longer effectively limits the rate of assembly7,
9. Unlike fluorescence, light scattering continues to increase at [NaCl]  1 M, probably owing to aggregation of kinetically trapped intermediates.
To better understand fluorescence quenching in C150BO assembly, we set up a series of assembly reactions containing the same total protein but different proportions of dye-labeled and unlabeled protein (Fig. 3a). Assembly results in little quenching in a 1:9 mixture of C150BO and C150. At this ratio, few capsid vertices have more than one fluorophore, eliminating intradimer quenching (cis quenching) during assembly. For higher proportions of dye-labeled molecules (1:4), quenching is enhanced (also see Supplementary Fig. 1 online). Thus, quenching is dominated by interactions between dye molecules conjugated to different dimers (trans quenching).
| | |
We hypothesize that quenching in trans is due to interactions between two or more BODIPY-FL molecules at a vertex (Fig. 3b). In the capsid, a dimer extends from a fivefold to a sixfold vertex or between two sixfolds, so that each dimer at a vertex will have four or five neighbors. We predict that some quenching will occur when at least one other dye-labeled subunit (gray) is among the subunits meeting the black subunit at a vertex. Thus, a ratio of greater than 1:4 labeled/unlabeled dimer is needed for substantial quenching. We observe quenching with the 1:4 mixture of labeled/unlabeled protein. If quenching could only occur between adjacent molecules, at least two labeled subunits in five would have to be labeled to induce quenching.
We tested the ability of our assay to detect compounds that alter HBV core-protein assembly in vitro by comparing fluorescence data to SEC data for assembly reactions containing either HAP-1 (ref. 10) or urea12. HAP-1 enhanced the rate and extent of HBV core protein assembly in vitro over a broad range of concentrations10. Higher HAP-1 concentration (10  M; that is, >1 HAP-1 molecule per dimer) led to the formation of aberrant particles owing to HAP-1's preference for capsid-protein hexamers rather than pentamers10. The decrease in fluorescence shows that HAP-1 substantially increased both the rate and the extent of C150BO assembly even at substoichiometric levels (Fig. 4a), as seen for the wild-type capsid protein.
| | |
Urea inhibits HBV capsid assembly at concentrations 0.75 M and causes reversible dissociation of wild-type HBV capsids at concentrations between 2.5 and 3.5 M without denaturing core-protein dimers12. We observed a decrease in assembly by SEC and an increase in fluorescence (loss of fluorescence quenching) in the presence of urea. C150BO still eluted as an apparent dimer by SEC at urea concentrations 0.75 M, demonstrating that the inhibition of assembly observed was not due to protein denaturation. We noted that low levels of HAP-1 or urea had intermediate effects on quenching.
To control for the possibility that either HAP-1 or urea was affecting BODIPY-FL fluorescence, we assayed the fluorescence of mock-assembled C150B0 dimer without NaCl. For HAP-1, these controls showed that the drug had no effect on fluorescence at early times (up to 2 h), even though the same amount of HAP-1 had a very strong effect on fluorescence of NaCl-induced assembly at the same times (data not shown). Therefore, the effect on fluorescence was due solely to increased assembly rather than to the drug quenching the fluorescence directly. After 24 h, the HAP-1 mock-assembled controls showed both BODIPY-FL quenching and assembly by SEC, consistent with our previous observation of slow assembly of wild-type capsid protein in the presence of HAP-1 under otherwise nonpermissive conditions10. In the mock-assembled controls, urea affected neither assembly nor quenching.
The BODIPY-FL fluorescence quenching assay was useful in identifying both misdirectors and inhibitors of HBV capsid-protein assembly. BODIPY-FL fluorescence was sensitive to subtle enhancement or inhibition, as well as more substantial effects. Using a 96-well format, we also tested a pool of potential small-molecule inhibitors. A weak misdirector was identified, matching the results we obtained by more labor-intense means (data not shown). We believe that this assay has potential in rapid screening for pharmaceutical lead compounds.
Exploitation of nonenzymatic aspects of the viral life cycle as therapeutic targets has begun only recently (see ref. 22 for review). Targeting processes such as virus entry, virion assembly and maturation, particle release and capsid uncoating is problematic because they are less readily assayed. Virus assembly is an integral part of the viral life cycle but has no similarity to processes in the uninfected cell23; thus it should provide a specific target for antiviral therapy. Very small changes in local conformation can cause massive changes in global capsid conformation, leading to aberrant products even though intersubunit interactions in each case are apparently very similar10. This conformational flexibility leaves assembly particularly susceptible to inhibition and/or misdirection.
Here we demonstrate a simple, rapid and sensitive fluorescence-based assay for HBV capsid assembly that is readily adaptable to a microtiter plate format and can be scaled up for high-throughput screening. The dynamic range of this assay allows us to detect both subtle and drastic effects on assembly (Figs. 1 and 4). This assay can be used to detect both inhibitors, which prevent or prematurely terminate capsid assembly, and misdirectors, which can hyperactivate capsid assembly leading to loss of regulation and/or the formation of aberrant products. An in vitro system has the advantage of rapidity (2–24 h as opposed to 5–7 d) and low cost compared to the currently available tissue culture model. An ex vivo system also enables identification of lead compounds otherwise rejected because of problems with uptake or toxicity, but which may be useful after chemical modification.
Assembly can be monitored in vitro by other techniques, but none of these can be conveniently performed on large numbers of replicate samples, and each has its own technical difficulties. For example, light scattering and turbidity measurements are extremely sensitive to particulates and large aggregates. Enzyme-linked immunosorbent assays are only possible for systems in which assembly-specific antibodies (such as anti-HBcAg) have been characterized. The fluorescence-quenching assay is rapid, works in a microtiter plate format, and is not sensitive to particulates, intermediates or aggregates. As molecular details of assembly are understood for more viruses, our approach may be adaptable to avian influenza24, hepatitis C virus25, HIV26 or any other viral system in which the accurate assembly of macromolecular complexes is essential for propagation or infection.
Methods
Mutagenesis.
For simplicity of chemical labeling, we mutated the cysteine residues (C48, C61 and C107) in the assembly domain of HBV strain adyw core protein (amino acids 1–149 (ref. 27), core protein 149) to alanine, inserting a unique cysteine residue at the C terminus (C150). Mutagenesis was performed using QuikChange Multi (Stratagene). Mutagenic primers are described in Supplementary Table 2 online (IDT DNA technologies). Mutations were confirmed by dye terminator dideoxy sequencing (DNA Sequencing Core, Oklahoma Medical Research Foundation). We refer to the resulting mutant protein as C150.
Protein expression, purification and dye labeling.
Wild-type and mutant truncated HBV capsid protein dimers were expressed and purified from E. coli as described27,
28. Protein was quantified by absorbance at 280 nm ( = 60,900 M-1 cm-1). DTT levels were maintained at 5 mM throughout purification and storage. Immediately before derivatization, C150 protein was removed from storage buffer by chromatography over a G25 PD10 desalting column (Amersham Biosciences) equilibrated with ice-cold 50 mM HEPES pH 7.5 without DTT. Peak fractions were reacted with BODIPY-FL maleimide (Invitrogen/Molecular Probes) on ice at a final concentration of 4 mM from a 20 mM BODIPY-FL stock in DMSO. Most complete labeling was achieved by overnight reaction, although significant labeling could be achieved by reaction for as little as 10 min. Unreacted dye was removed by separation over a G25 PD10 desalting column equilibrated in ice-cold 50 mM HEPES, pH 7.5. The degree of dye labeling was determined using absorbance at 504 nm (504 = 73,000 M-1cm-1, 280 = 1,300 M-1cm-1). The extinction coefficient was calculated by determining the degree of labeling of BODIPY-FL-labeled C150 preps by MALDI-TOF mass spectrometry. Yield of BODIPY-FL-labeled C150 was typically on the order of 1.9 moles of dye per mole of dimer. Labeling of C150 with fluorescein maleimide was compared to BODIPY-FL; BODIPY-FL conjugates gave a much stronger assembly-dependent change in fluorescence than did fluorescein. Random labeling of  and amino groups with succinimidyl conjugates was inefficient, and the labeled proteins showed little or no change in fluorescence on assembly.
Fluorescence and light scattering.
Assembly was monitored by fluorescence using a SPEX Fluoromax-2 fluorometer (Horiba Jobin Yvon), using a 0.3-cm path-length cuvette (Hellma) or in black 96-well COSTAR fluorescence microtiter plates (Corning) using a MicroMax adaptor. Assembly reactions were performed at 21 °C. Assembly was initiated manually by mixing core protein in 50 mM HEPES, pH 7.5, with buffered NaCl (Sigma) as appropriate. Fluorescence was excited at 504 nm and emission was measured at 509 nm (1-nm band pass for each). For static measurements, all measurements were recorded in triplicate. Qualitatively similar results were obtained using excitation at 495 nm and emission at 512 nm (commonly used for fluorescein fluorescence). Measuring fluorescence of the same reactions in either a cuvette or a 96-well fluorescence microtiter plate gave identical quenching values.
Where assembly was monitored by 90° light scattering, excitation and emission were set at 400 nm with a 3-nm band pass using a 0.3-cm path length cuvette as previously described7,
10. Light scattering was measured at 400 nm, rather than 320 nm as previously, to minimize the effect of absorbance of both HAP-1 and BODIPY-FL at the shorter wavelength. The intense scattered light was attenuated by a neutral density filter. Light scattering was monitored in real time during an assembly reaction, or determined in triplicate after 24-h incubation under assembly conditions.
Size-exclusion chromatography.
Assembly reactions were examined by SEC on a Superose 6 10/30 column (Amersham Biosciences) mounted on a Shimadzu high-performance liquid chromatography system equipped with an auto injection module (Shimadzu). The column was equilibrated with 50 mM HEPES, pH 7.5, 50 mM NaCl. Recovered protein was assigned either to the void (6.5–7.0 ml), capsid (7.0–8.3 ml), dimer (15–16.5 ml) or intermediate elution (8.3–15 ml).
Note: Supplementary information is available on the Nature Biotechnology website.
Received 8 August 2005; Accepted 12 December 2005; Published online: 12 February 2006.
REFERENCES
-
Previsani, N.
&
Lavanchy, D.
Hepatitis B (Epidemic and Pandemic Alert and Response, World Health Organization, Geneva, 2002). http://www.who.int/csr/disease/hepatitis/whocdscsrlyo20022/en/
-
Ganem, D.
Hepadnaviridae and their replication. in Fields virology (eds. Fields, B.N. et al.) 2703–2737, (Lippincott–Raven Publishers, Philadelphia, 1996).
-
Summers, J.
&
Mason, W.S.
Replication of the genome of a hepatitis B-like virus by reverse transcription of an RNA intermediate. Cell 29, 403–415 (1982). | Article | PubMed | ISI | ChemPort |
-
Wingfield, P.T.
,
Stahl, S.J.
,
Williams, R.W.
&
Steven, A.C.
Hepatitis core antigen produced in Escherichia coli: subunit composition, conformational analysis, and in vitro capsid assembly. Biochemistry 34, 4919–4932 (1995). | Article | PubMed | ISI | ChemPort |
-
Nassal, M.
,
Rieger, A.
&
Steinau, O.
Topological analysis of the hepatitis B virus core particle by cysteine-cysteine cross-linking. J. Mol. Biol. 225, 1013–1025 (1992). | Article | PubMed | ISI | ChemPort |
-
Zhou, S.
&
Standring, D.N.
Hepatitis B virus capsid particles are assembled from core-protein dimer precursors. Proc. Natl. Acad. Sci. USA 89, 10046–10050 (1992). | PubMed | ChemPort |
-
Zlotnick, A.
,
Johnson, J.M.
,
Wingfield, P.W.
,
Stahl, S.J.
&
Endres, D.
A theoretical model successfully identifies features of hepatitis B virus capsid assembly. Biochemistry 38, 14644–14652 (1999). | Article | PubMed | ISI | ChemPort |
-
Ceres, P.
&
Zlotnick, A.
Weak protein-protein interactions are sufficient to drive assembly of hepatitis B virus capsids. Biochemistry 41, 11525–11531 (2002). | Article | PubMed | ISI | ChemPort |
-
Stray, S.
,
Ceres, P.
&
Zlotnick, A.
Zinc ions trigger conformational change and oligomerization of hepatitis B virus capsid protein. Biochemistry 43, 9989–9998 (2004). | Article | PubMed | ISI | ChemPort |
-
Stray, S.
et al. A heteroaryldihydropyrimidine activates and can misdirect hepatitis B virus capsid assembly. Proc. Natl. Acad. Sci. USA 102, 8138–8143 (2005). | Article | PubMed | ChemPort |
-
Zlotnick, A.
Are weak protein–protein interactions the general rule in capsid assembly? Virology 315, 269–274 (2003). | Article | PubMed | ISI | ChemPort |
-
Singh, S.
&
Zlotnick, A.
Observed hysteresis of virus capsid disassembly is implicit in fundamental kinetic models of association and dissociation. J. Biol. Chem. 278, 18249–18255 (2003). | Article | PubMed | ISI | ChemPort |
-
Weber, O.
et al. Inhibition of human hepatitis B (HBV) by a novel non-nucleosidic compound in a transgenic mouse model. Antiviral Res. 54, 69–78 (2002). | Article | PubMed | ISI | ChemPort |
-
Deres, K.
et al. Inhibition of hepatitis B virus replication by drug induced depletion of nucleocapsids. Science 299, 893–896 (2003). | Article | PubMed | ISI | ChemPort |
-
Hacker, H.
,
Deres, K.
,
Mildenberger, M.
&
Schroder, C.
Antivirals interacting with hepatitis B virus core protein and core mutations may misdirect capsid assembly in a similar fashion. Biochem. Pharmacol. 66, 2273–2279 (2003). | Article | PubMed | ISI | ChemPort |
-
Zlotnick, A.
et al. Localization of the C terminus of the assembly domain of hepatitis B virus capsid protein: implications for morphogenesis and organization of encapsidated RNA. Proc. Natl. Acad. Sci. USA 94, 9556–9561 (1997). | Article | PubMed | ChemPort |
-
Wynne, S.A.
,
Crowther, R.A.
&
Leslie, A.G.
The crystal structure of the human hepatitis B virus capsid. Mol. Cell 3, 771–780 (1999). | Article | PubMed | ISI | ChemPort |
-
Nassal, M.
Conserved cysteines of the hepatitis B virus core protein are not required for assembly of replication-competent core particles nor for their envelopment. Virology 190, 499–505 (1992). | Article | PubMed | ISI | ChemPort |
-
Bergström, F.
et al. Dimers of dipyrrometheneboron difluoride (BODIPY) with light spectroscopic applications in chemistry and biology. J. Am. Chem. Soc. 124, 196–204 (2002). | PubMed |
-
Johnson, I.
,
Kang, H.
&
Haugland, R.
Fluorescent probes incorporating dipyrrometheneboron difluoride fluorophores. Anal. Biochem. 198, 228–237 (1991). | Article | PubMed | ISI | ChemPort |
-
Ceres, P.
,
Stray, S.
&
Zlotnick, A.
Hepatitis B virus capsid assembly is enhanced by naturally occurring mutation F97L. J. Virol. 78, 9538–9543 (2004). | Article | PubMed | ISI | ChemPort |
-
de Clercq, E.
Antivirals and antiviral strategies. Nat. Rev. Microbiol. 2, 704–720 (2004). | Article | PubMed | ISI | ChemPort |
-
Zlotnick, A.
&
Stray, S.
How does your virus grow? Understanding and interfering with virus assembly. Trends Biotechnol. 21, 536–542 (2003). | Article | PubMed | ISI | ChemPort |
-
Harris, A.
,
Forouhar, F.
,
Qiu, S.
,
Sha, B.
&
Luo, M.
The crystal structure of the influenza matrix protein M1 at neutral pH: M1-M1 protein interfaces can rotate in the oligomeric structures of M1. Virology 289, 34–44 (2001). | Article | PubMed | ISI | ChemPort |
-
Kunkel, M.
,
Lorinczi, M.
,
Rijnbrand, R.
,
Lemon, S.
&
Watowich, S.
Self-assembly of nucleocapsid-like particles from recombinant hepatitis C virus core protein. J. Virol. 75, 2119–2129 (2001). | Article | PubMed | ISI | ChemPort |
-
Campbell, S.
&
Vogt, V.
Self-assembly in vitro of purified CA-NC proteins from Rous sarcoma virus and human immunodeficiency virus type 1. J. Virol. 69, 6487–6497 (1995). | PubMed | ISI | ChemPort |
-
Zlotnick, A.
et al. Dimorphism of hepatitis B virus capsids is strongly influenced by the C-terminus of the capsid protein. Biochemistry 35, 7412–7421 (1996). | Article | PubMed | ISI | ChemPort |
-
Zlotnick, A.
,
Ceres, P.
,
Singh, S.
&
Johnson, J.M.
A small molecule inhibits and misdirects assembly of hepatitis B virus capsids. J. Virol. 76, 4848–4854 (2002). | Article | PubMed | ISI | ChemPort |
-
Zlotnick, A.
Theoretical aspects of virus capsid assembly. J. Mol. Recognit. 18, 479–490 (2005). | Article | PubMed | ISI | ChemPort |
Acknowledgments
This work was supported by a Research Scholar Grant (RSG-99-339-04-MBC) from the American Cancer Society and National Institutes of Health Grant R01-AI067417. We thank Sheryl Christofferson of the Oklahoma Medical Research Foundation Sequencing Core Facility, Bruce Baggenstoss of the OUHSC EPSCOR Mass Spectrometry Facility, Pablo Ceres for assistance with design of mutagenic primers, Christina Bourne for distance determinations from structural data and Laura Buford, Brian Firestone and Quincie Phan for excellent technical assistance. We thank Gillian Air for critical reading of the manuscript.
Competing interests statement:
The authors declare competing financial interests.
|
standard deviation of three replicates, including samples from 
